Perplexity AI targets 1 mn user sign-ups by mid-March, may offer free plan

Aravind Srinivas plans to make its professional service plan free for students in India

Perplexity AI, an artificial intelligence (AI) search startup, is targeting a million user signups by mid-March, hinting that its premium service plan may be offered for free to students in the country.

 

The US-based company is betting on specialising in trained models that can achieve improved outputs alongside improved experience for its users over pre-training models from datasets.

 

“We’re working with a student group on WhatsApp and aiming for close to a million signups by mid-March. Our goal is to make sure no student has to pay for the pro version of Perplexity anywhere in the world. We do want to bring it to everybody in India, whoever is a student,” said Aravind Srinivas, co-founder and chief executive officer (CEO), Perplexity.

 

At present, the professional plan at the company costs $20 per month despite a user’s geographical location. The search service continues to be free with its standard offering.

 

He was speaking alongside Paytm founder Vijay Shekhar Sharma at a virtual launch event of the fintech company’s partnership with the AI search company.

 

“A lot of factors beyond just core model building will matter and our bet has always been that post-training (models) and RL (reinforcement learning) matters more than even pre-training. Pre-training will be a commodity and we want to specialise more in post-training and user experience,” he added.

 

The race to build AI models and deploy tools for users is heating up among major global firms, including OpenAI’s ChatGPT, Google’s Gemini, Meta’s Llama, Anthropic, among others. Srinivas explained that most AI models will end up with similar capabilities over time.

 

“Models are always going to be awesome incrementally, but they all end up looking very similar in terms of capabilities because there are only a few core IQ-point-related benchmarks that you can measure model capability by,” he said.

 

He added that differentiation will come from how companies harness reasoning, summarisation, and synthesis capabilities of these models for building a refined product experience.

 

On biases, Srinivas elucidated that a way to address the challenge would be to cover a diverse set of data sources and retrieve as many different opinions as possible.